The Impact of Design Rework on Construction Project
Performance
Ying Li
Graduate Student
University of Kentucky, College of Engineering
Department of Civil Engineering, 116 Raymond Building, Lexington, KY40506-0281, USA
(859) 257-1036
yli239 @uky.edu
Timothy R. B. Taylor
Assistant Professor
University of Kentucky, College of Engineering
Department of Civil Engineering, 151A Raymond Building, Lexington, KY40506-0281, USA
(859) 323-3680
taylor @engr.uky.edu
Abstract: Rework in construction development projects can significantly degrade project cost
and schedule performance. In a typical construction development project which involves design
and construction, rework in the construction phase could increase construction cost by
10%-15% of the contract price (Burati. 1992, Josephson & Hammerlund 1999, Love & Li 2000).
The proportion of money and time spent on rework in the design phase is usually higher (Smith
& Eppinger 1997). In large, complex projects, undiscovered rework in the design phase can
induce rework in the construction phase. The time when rework is discovered during the project
development process affects the impact of rework on overall project performance. However,
available knowledge is not always successful in improving project managers’ understanding of
the feedback mechanisms which drive undiscovered rework impacts on project performance,
specifically the interaction between different phases during the developing process. The current
work uses a system dynamics model of a two phase project development cycle to identify high
leverage points for minimizing the impacts of rework on development project performance.
Model analysis suggests that failing to discover rework near its creation in the project
development process can magnify the impact of rework on project performance.
1 of 15
1. Introduction
Rework is a common occurrence in construction projects and has been identified as one of the
factors that can degrade project performance. Over the years researchers have developed
definitions and interpretations of rework in correspondence to their own production systems.
Love (2000) defines rework in the construction industry as the “unnecessary effort of redoing a
process or activity that was incorrectly implemented the first time.” The Construction Industry
Institute (CII) defines field rework as “‘activities that have to be done more than once or activities
that remove work previously installed as part of a project” (CII 2002). Rework in development
projects can significantly degrade project cost and schedule performance. Research shows that
rework in the construction phase could increase costs by 4% to 12% of the construction contract
amount (Burati. 1992; Josephson & Hammerlund 1999; Love 2000). The proportion of money
and time spent on rework in the design phase is usually higher than that of the construction phase,
as design is an iterative process during which engineers try to solve coupled problems with
complex relationships (Smith & Eppinger 1997). Sometimes design tasks are so closely related
that each task, if not completed perfectly, has a probability of creating rework for another task.
Under the pressure to improve project cost and schedule performance, many companies have
accepted the fast-tracking approach under which the design phase and the construction phase
overlap (Pefia-Mora and Li 2001; Fazio, Moselhi, Théberge and Revay 1988). Because of this
phase overlap it is possible that a contractor can start the construction phase with flawed plans
that have undiscovered errors (referred to as “design undiscovered rework” in the current work).
In large, complex projects undiscovered rework in the design phase can produce a significant
amount of rework in the construction phase.
>
2. Problem Description
When rework occurs during the development process, a project can experience poor cost and
schedule performance. In that case, the project manager’s attention is focused on completing
work faster with limited resources, which creates a “fire-fighting” situation. In the product
development context, fire fighting describes “the unplanned allocation of engineers and other
resources to fix problems discovered late in a product's development cycle” (Repenning, 2001).
When fire fighting begins, the project managers may focus their attention on completing project
scope or addressing known rework. This reduction in attention to quality assurance may result in
a reduction in quality assurance effectiveness which would result in some errors going
undetected and being approved as correctly completed work. Defective work that is approved is
referred to hereafter as undiscovered rework. In traditional construction project delivery systems,
the project development process consists of a design and construction phase. Undiscovered
rework created during the design phase can create additional work in the construction phase.
Two types of additional work may be created by undiscovered rework. One is work that was not
in the initial project scope but has to be completed to support those parts of the project that are
related to the part being reworked. Creation of this type of additional work is sometimes referred
to as “ripple effects” and implies that additional work beyond the initial project scope is needed
(i.e. new work is added to the project). The other type is work that was in the initial project scope
and was initially installed correctly but needs to be reworked because it’s closely related to a
separate item of work being removed. Consider the situation in which an engineer designing a
roadway project made an error sizing storm sewer pipes that pass under the roadway and this
2 of 15
error was not identified by the design quality assurance process. During construction the wrong
size pipes were installed underneath the pavement and the error in pipe sizing is not discovered
until after the pavement covering the pipes has been placed. In order to correct the pipe sizing
error (the undiscovered design rework), the pavement above the pipe must be demolished (work
that was completed correctly but required rework due to rework in adjacent systems) and the
excavation must be shored (work that was not required as part of the original project scope).
In complex projects where activities are closely related to each other, the longer it takes to find
the mistake, the more additional work can be created in the process of correcting the mistake and
the more the total project performance can be degraded. In the previous example, had the pipe
sizing error been discovered during the design phase, additional design time would be required to
correct the rework but this would have been much less than the cost of replacing the pipes after
installation. However, in the traditional design-bid-build construction process total project cost is
determined by the summation of the design costs (managed by the designer) and construction
costs (managed by the contractor). Discovering design errors during the design phase can
decrease the overall project costs but it could also increase the design cost which works against
the profit motive of the designer. Failing to discover design errors during the design phase
decreases the design costs while increasing the construction cost (which works against the profit
motives of the contractor). The project owner is concerned with the entire project cost but does
not have direct management of the design and construction costs. How can undiscovered rework
be best managed under the feedback dynamics of the design and construction process?
Previous research has examined undiscovered rework (Ford and Sterman 1998) and rework
induced ripple effects (Taylor & Ford, 2006, 2008) in single phase project development
processes. The current work extends this research by investigating the impact of undiscovered
rework on project performance in the two phased construction project development process. The
objectives of this research are: (1) identify the feedback mechanisms that drive the behaviour of
undiscovered rework in the two phased construction project development process; (2) identify
the impacts of undiscovered rework on project cost and schedule performance within the two
phased construction project development process; (3) design and test policies for managing
rework in the two phased construction project development process.
3. A Simulation Model of Project Developing Process
System dynamics is a methodology for studying and managing complex systems (Sterman,
2000). System dynamics models have been successfully applied to project management issues
including the effect of rework on project performance (Cooper 1994), tipping point dynamics
(Taylor and Ford 2006, 2008), failures in fast track implementations (Ford and Sterman 1998).
The system dynamics methodology was selected for the current work because it clearly
illustrates the dynamics of design error induced rework in the two-phased construction project
development process. The system dynamics model used in the current work is based on the
structure of Taylor and Ford’s (2006, 2008) tipping point model. The current work expands this
model by using the tipping point structure to model undiscovered rework in the two phase
construction project development project. The model consists of three sectors: A work flow
sector, a resource allocation sector, and a cost accounting sector. Figure | is a simplified version
of the work flow sector. The model contains a design phase and a construction phase. For each of
3 of 15
the two phases, the model uses the basic workflow structure used in the Taylor and Ford’s (2008)
model. The “rework fraction” variables indicate the possibility that a work package is completed
incorrectly during initial completion. The “Design QA effectiveness” variables indicate the
possibility that design quality assurance staff catch an error and send it to rework backlog.
The feedback loops identified in Figure 1 (R1) can be used to describe the impact of having
undiscovered rework on the project performance within the same phase. When quality assurance
is not 100% effective, mistakes are not identified by quality assurance and are sent to the
“Undiscovered Rework Backlog.” However, since the errors were not identified, the project
manager perceives that this “undiscovered rework backlog” as being included in the stock of
“Work Released.” This undiscovered rework can create additional rework. For example, having
installed underground water pipes of the wrong size without knowing it until the end of the
project may create rework in foundation, masonry, electrics, painting, etc. Since having
undiscovered rework can create more rework, it increases the rework fraction of the same phase
which increases the number of errors. With the same quality assurance effectiveness, more
mistakes will be sent to “Undiscovered Rework Backlog” which will further increase the rework
fraction.
The stock-flow structure in the construction phase is slightly different from that of the design
phase. For simplicity the current model assumes that that construction quality assurance is 100%
effective (i.e. all errors are discovered) so that errors made in the construction phase which
resulted from design undiscovered rework can be tracked and therefore the impact of design
undiscovered rework on construction performance is captured.
The two phases are related according to the following logic. The construction phase only starts
when the design reaches a certain percent complete, which is determined by the project
management team. Having undiscovered rework in the design phase could also increase the
rework fraction in the construction phase if the design error is not identified before it is
implemented in the construction phase. When the construction staff identify a design error, the
design staff must rework the work package. In Figure 1, this situation is reflected by the “Design
discover undiscovered rework rate.” This flow moves work packages from the “Design
Undiscovered Rework Backlog” to “Design Known Undiscovered Rework Backlog” as
construction staff finds design errors. The bold arrow in Figure | corresponds to this feedback
from construction phase to design phase. The “Design Known Undiscovered Rework Backlog”
stores undiscovered design rework identified by construction staff but have not been corrected by
the design staff. As the design staff fix the error, the relevant work package will then be
re-released in the construction phase, through the “Construction receive corrected design rate” in
Figure 1. The bold dashed arrow in Figure | corresponds to this feedback from design phase to
construction phase.
4 of 15
Figure 1 Simplified Model Structure (Work Flow Sector Only)
Design Phase Construction Phase
<Design known
undiscovered rework
complete rate>
? °
‘Construction
Rework due to|
construction
olikdiscover design
} UR rate
+ os °
‘Construction fraction FZ *
discovered to require
change og
‘
Construction base
fraction disc to g-——h
require reworks, ° ‘Construction
Design base
fraction disc to
require rework,
esign fraction
discovered to
5 of 15
Backlog Xe QArate o
Design i S40f design undiscovers Construction
renting discovery ~ YSN on sin weal discover rom
o rate yg released to constructior rate
- ib.
Design ‘ Frage re a ease
8 ality
Design approve [Released <Pesign Assurance jon LReleased
werk rate 2 inci Bacldag® | Construction
Design add new i mi me Bene 9: work rate 5
tasks rate erceived design a
o percent complete, Carter
add new tasks
rate
‘ompl
work rate
°
° Minimum design
percent complete to
start construction |
In Taylor and Ford’s 2008 model, resource (manpower) is allocated to all the backlogs in direct
proportion to the amount of work packages in each backlog. The structure of the resource
allocation sector in the current model is similar to the Taylor and Ford’s 2008 model with the
design and construction phases having separate resource allocation sectors.
In the design phase, the cost is determined by the total number of tasks the designers perform,
which is an indicator of the size of the project. The model assumes that an experienced project
manager would be able to forecast the cost of rework according to past practice before the
project starts and include the extra effort in their initial budget. So if the original scope is made
up of n work packages, each work package has to go through two (2) tasks (initial completion
and quality assurance) to get released, each task costs A dollars, and the estimated rework
fraction (r) is 0.2, the initial project budget will be:
Initial Budget = cna (1)
As the project proceeds, the project manager tracks project cost to date and estimates the money
required to complete the rest of the project. The method is to check the number of work packages
in each of the backlog in progress, i.e. the “Initial Completion Backlog”, the “Quality Assurance
Backlog”, the “Rework Backlog” and the “Known Undiscovered Rework Backlog” in the design
phase. The project manager then estimates the number of tasks needed for one work package in
each of these backlogs, and apply task unit cost to get the money needed to complete the rest of
the project. For example, for each work packages in the “Initial Completion Backlog”, if the
estimated rework fraction is r, then the chance that it will take two (2) tasks (initial completion
and quality assurance) to complete this work package is (1-r). The chance that it will take four (4)
tasks (initial completion, quality assurance, rework and quality assurance again) to complete this
work package is r * (1-r). The chance that it will take six (6) tasks (initial completion, quality
assurance, rework, quality assurance, rework again, and quality assurance again) to complete this
work package is r° * (1-r), and so on. Therefore the expected number of tasks needed for each
work package in the initial completion backlog is:
22irta-n)= 4 2)
Similarly, the number of tasks needed for each work package in the rework backlog and the
known undiscovered rework backlog is:
22irta-nj= 4 (3)
6 of 15
And the number of tasks needed for each work package in the quality assurance backlog is:
2 ((2i- pret -r)] = 14+4— (4)
1-r
Construction labor cost and non-labor cost are calculated separately. The same structure as
described above is used to track construction non-labor cost. Construction labor cost is calculated
according to wage rates, crew sizes and construction duration (See Figure 2).
Figure 2 Construction Labor Cost
construction
wage rate <Total
construction staff,
OS Construction
add to ee Duration
duration ;
construction
a of labor cost
3
Perceived a done>
percent complete>
<Minimum design percent <Construction construction cost
complete to start percent complete>, todate
construction>
<Construction-Non-Labor
Money Used>
The total project costs are then determined using the following equation.
Total Project Cost = Design Cost + Construction Cost (5)
When the project experiences cost overrun, the project team is running in an “under-resourced”
condition, which causes the “Design QA effectiveness” to decrease as the management team
focuses their attention on other issues.
4. Typical Model Behaviour, Testing, and Analysis
The model was tested using standard system dynamics procedures (Sterman 2000). The model is
able to reflect situations and comparison of different conditions as expected. For example, when
design quality assurance effectiveness is 1 (perfect quality assurance), no work is placed in the
“Design Undiscovered Rework Backlog” and the “Construction Rework Due to Design
Undiscovered Rework” backlog. When quality assurance effectiveness drops, undiscovered
rework is placed in the “Design Undiscovered Rework Backlog” and the lower the design quality
assurance effectiveness, the more undiscovered rework is created. When quality assurance
effectiveness drops, both cost performance and schedule performance degrade. Figure 3 shows
significant degrade in cost performance and schedule performance when design quality
assurance effectiveness is 80% and 50% as compared to 100%.
7 of 15
Figure 3 Comparison of Project Cost Performance and schedule performance
Total Project Cost to Date Project Duration
4M 80
« 2M #z i 40
a
0 4 0
0 9 18 27 36 45 54 63 72 81 90 99 0 18 36 54 72 90
Time (Week) Time (Week)
Desi
Du
Des
To further test if all the major control variables impact the system behaviour in reasonable ways,
we performed statistical screening test (Taylor et al, 2010) on six (6) major control variables
regarding the correlation coefficient between them and the total project cost to date. Statistical
screening is a simple, structured, and user-friendly method of identifying high-leverage model
parameters. It uses multiple simulations generated by varying model input parameters to
calculate linear correlation coefficients that measure the direction and strength of the relationship
between input parameters and a user-defined system performance variable. Since we are most
interested in project cost performance, we selected total project cost to date as the performance
variable. When choosing control variables, we selected the exogenous parameters that describe
the interaction of the two design phases and the rework fractions in both phases, since they
indicate the project’s level of complexity and are related to a great number of other variables in
the model. We assigned an uncertainty of +25% from the base case value for each control
variable and varied the parameter values according to a uniform distribution. This range is
determined by taking both reality and rationality into consideration. For example, the base case
value for design quality assurance effectiveness is 0.6 and the value used for each simulation
can’t exceed 1 because the quality assurance staff can’t find more errors than there actually are.
In Figure 4, if the correlation coefficient is between -0.2 and 0.2, the correspondent control
variable is not significantly related to the performance variable (Taylor et al, 2010). If a
correlation coefficient is above 0.2, the polarity between the control variable and the
performance variable is positive, which means the control variable and the performance variable
move in the same direction (i.e. an increase in the input causes an increase in the model output
and vice versa). The higher the correlation coefficient is, the stronger the relationship. If a line
goes below -0.2, the polarity between the control variable and the performance variable is
negative, which means the control variable and the performance variable move in the opposite
direction (i.e. an increase in the input parameter causes a decrease in the model output and vice
versa). The lower the coefficient is, the stronger the relationship.
8 of 15
Figure 4 Correlation Coefficients-Total Cost to Date and Construction Percent Complete
Correlation Coefficients
Correlation Coeftecients
Correlation Coefficients-Total Cost to Date
1
6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Correlation Coefficients- Percent Complete
1
6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Time
—— Construction base fraction disc to require rework
Design base fraction discto require rework
—+— Design Base QA Effectiveness
Design Sensitivity of rework fraction to undiscovered rework
—--—: Minimum design percent complete to start construction
~~~ Sensitivity of construction rework fractionto design undiscovered rework
The test results show that with this model structure, increasing the design rework fraction or
construction rework fraction will increase total project cost and cause significant delay in project
schedule. The design rework fraction has a slightly stronger influence to both performance
variables, since design rework impacts both phases. Improving the design base quality assurance
9 of 15
effectiveness will slightly improve cost performance. The earlier construction phase is started,
the greater the project cost given the increased potential for rework.
To identify high leverage points that managers can use to improve project performance, a
sensitivity analysis was performed on the six (6) major control variables in the model, and results
show that design rework fraction, construction rework fraction and design quality assurance
effectiveness have the greatest impact on project cost performance. While it can be difficult for a
project manager to reduce the rework fraction, design quality assurance is more within the
project manager’s control. Figure 5 below shows the overall project cost when design quality
assurance effectiveness, design base rework fraction and construction base rework fraction vary
from -25% to +25% as compared to their base case values. In the base case, the design rework
fraction and construction rework fraction are set to be 0.2, and design base quality assurance
effectiveness is 0.8. The relationship between total project cost and design base quality assurance
effectiveness and that between project duration and design base quality assurance effectiveness
appear to be linear within this interval, while the design base rework fraction line and the
construction base rework fraction line resemble an exponential growth curve.
Figure 4 Sensitivity Analysis (only showing the three most influential variables)
Total Project Cost
$3,100,000
$2,900,000 =
$2,700,000 wane — - Design Base Rework
$2,500,000 ===. i Fraction
$2,300,000 -ps=, ase ————_ = = = Construction Base
$2,100,000 . Rework Fraction
: ' Design Base QA
$1,900,000 Effectiveness
$1,700,000
$1,500,000 T :
-0.25 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25
Project Duration
75
70
65 —-- 7 — - Design Base
SS Rework Fraction
¥ 60 —=
o Lager === Construction Base
5S 55 Rework Fraction
50 Design Base QA
45 Effectiveness
40
0.25 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 02 0.25
10 of 15
To further investigate the relationship between the two most influential variables and project
performance variables, a sensitivity analysis with wider intervals on these two variables were
performed. Results are shown in Figure 5. For each line in Figure 5, while the indicated control
variable varies, the other control variable remains its base case value (0.2). For projects that are
simple (those have low design rework fraction and construction rework fraction), the
development process only generates limited number of rework and undiscovered design errors,
and the performance variables are more sensitive to construction base rework fraction, since
construction task unit cost is much greater (three times in the model) than design task unit cost
and it also takes longer to complete a task in construction than in design. As the two control
variables move into the higher range, the model becomes more and more sensitive to the design
base rework fraction. High design rework fraction usually means that the project is unique or
very complex. Model analysis shows that for this type of projects, the manager should pay
particular attention to minimizing the design rework fraction and preventing design errors from
entering the construction phase.
Figure 5 Sensitivity Analysis on Design Base Rework Fraction and Construction Base Rework
Fraction
Total Project Cost Project Duration
2.808+06 5
2.60+06
2.40E+06
Week
“1 s0e+0s
1.60E+06
1.40E+06 } 40
1.20E+06 35
1.008+06 30
0.02
0.04
0.06
0.08
O41
0.12
0.14
0.16
0.18
0.2
0.22
0.24
Design base fraction disc to require rework
— — —Construction base fraction disc to require rework
The statistical screening and sensitivity analysis results suggests that failing to discover rework
near its creation results in more rework which degrades project cost performance. In the base
case run, the design quality assurance effectiveness starts at 80%, and then drops as the project
experiences cost overruns. When compared to a project with the same amount of base rework but
a 100% effective quality assurance, the base case results in a 22% cost overrun compared to the
effective quality assurance policy.
11 of 15
5. Project Management with Design Rework
Although the design rework fraction and construction rework fraction are the two most related
variables to project performance, they are difficult to reduce by project manager. They depend in
some part on the level of complexity of the project and the competency of the project team.
However, there are strategies that a project manager can use to improve design quality assurance
effectiveness or at least eliminate the impact of design undiscovered rework on the construction
phase since the construction costs typically represent the majority of the total project cost in a
real construction project.
One obvious theoretical solution to solve the problem caused by design undiscovered rework is
to implement a design quality assurance program that identifies all design errors (i.e. 100%
quality assurance effectiveness) (Policy 1 in Table 1). Simulated results show that this policy
could lead to a 5.86% improvement in design cost, a 21.82% improvement in construction cost,
and a 17.72% improvement in total cost over the base case. For a project manager, keeping
design quality assurance perfect means never compromise quality assurance under any
conditions. Even when the project team is experiencing budget and schedule pressure, the project
manager must still provide sufficient resources and competent personnel to the quality assurance
department. What is most interesting with Policy 1 is that due to the structure of the two phase
design process and the increased cost of construction compared to design in the model the
Contractor receives a greater benefit of adopting Policy 1. However, the Contractor does nothing
to help solve the problem in this scenario (i.e. the contractor does not participate in the design
process).
Table 1 Comparison of improvement by using two policies.
Base Case Policy 1 Policy 2
Design staff finds Construction staff
1 afi With the problem of design UR before finds design UR
Description A inane : a
undiscovered rework | construction (design before construction
QA perfect) initial completion
| ees $2,416,000 $1,988,000 $2,162,000
% Improved - 17.72% 10.51%
Design Cost ($) $621,290 $584,876 $559,135
% Improved - 5.86% 10.00%
o cor $1,794,710 $1,403,124 $1,603,865
% Improved - 21.82% 10.63%
Project Duration 63 54 54
(week)
% Improved - 14.29% 14.29%
12 of 15
The problem can be solved by the construction firm, too. Knowing that design quality assurance
is not always perfect, the construction project manager can improve construction performance by
eliminating the impact of design undiscovered rework on the construction phase. For example,
they can hire a couple of engineers to recheck the drawings that were approved by design quality
assurance department prior to initial completion in the construction phase (Policy 2 in Table 1).
By doing this, most of the design errors can be found before the work is installed. Since the
design firm is doing less work than in the scenario of Policy 1, they will enjoy lower cost (a 9.33%
improvement in design cost compared to 5.86% of Policy 1). But the Contractor needs to pay
extra money to the engineers they hired to recheck the drawings. When taking the extra cost into
consideration, Policy | is clearly more attractive than Policy 2 for the Contractor. This highlights
a structure problem inherent in the traditional design-bid-build process. Both the Engineer and
the Contractor want to improve their own performance, but neither of them would want to
perform extra work for less improvement in return. Then balancing effort and reward will be up
to the Owner’s project management team to coordinate the Engineer and the Contractor to
achieve the lowest overall project cost for the owner.
The above comparison of policies is based on the assumption that the cost for checking the
drawings in the construction phase equals the cost of checking them in the design phase (design
quality assurance unit cost). But this may not be the case in real practice, and the difference in
costs for checking the drawings may alter the preference between the two policies. Therefore,
another sensitivity analysis was conducted on the difference of costs for checking the drawings
in the design phase and in the construction phase. The control variable is the unit cost for
checking drawings in the construction phase with the unit cost of checking drawings in the
design phase being $1,000. The results are shown in Figure 6.
Figure 6 Sensitivity Analysis on Checking Drawings Cost between Two Phases
Total Project Cost
$3,500,000
$3,000,000
$2,500,000
uw
$2,000,000
$1,500,000
$1,000,000
100 600 1100 1600 2100 2600 3100 3600 4100 4600
$
construction check drawings unit cost
13 of 15
Referring to Table 1, the total project cost for adopting Policy 1 is $1,988,000. If Policy 2 is used,
the total project cost is less than Policy 1 only when the cost of checking the drawings in the
construction phase is less than half of the cost of checking the drawings in the design phase.
Otherwise Policy 1 will be the preferred policy.
6. Conclusions and Implications for Practice
Project cost and schedule performance are controlled by the feedback mechanisms in the
development process and the current work shows that the rework cycle can alter project
performance. Looking at the whole entire two-phase construction development process, it is
always good to realize and fix the rework near the point of rework creation. When more than one
phases are included in the developing process, which is always the case for a construction
project, the errors missed by quality assurance staff and therefore released from the preceding
phase can have an great impact on the performance of the following phase. By identifying high
leverage points using the model described in this paper, the current work shows that having
undiscovered rework in the design phase will result in a significant cost overrun in the
construction phase. Both the design phase and the construction phase will benefit from
eliminating the impact of design undiscovered rework. First, all the parties involved in the
project must understand the potential result of having design undiscovered rework, then they
must work together toward the same goal, i.e. the ultimate project performance, in the meanwhile
balancing effort and reward of each party.
The current work offers several contributions to the existing body of knowledge. The proposed
model structure provides a structured feedback description of how design undiscovered rework
impact project performance in both the design phase and the construction phase, as well as
evaluates possible solutions to the problem of interest. This model structure is based on previous
accepted model, so that is can be calibrated to be suitable for typical development projects. The
added structure offers a view into the interaction between the design phase and the construction
phase from a system dynamic perspective.
The current work has several limitations which can be addressed in future work. The proposed
model structure mostly focuses on the impact of design phase on the construction. There are
possible feedbacks from the construction phase to the design phase as well, but are not
recognized in our model. For example, some construction mistakes could be better handled by
modifying design, rather than tearing the work down and reinstalling according to the original
design. The model work can also be expanded to test the effectiveness of other strategies in
managing the design-construction process (i.e. design-build). Finally, the model can be expanded
to test the impact of undiscovered design rework on tipping point dynamics (Taylor and Ford
2006, 2008) in the design and construction process.
14 of 15
7. References
Burati, J. J. 1992. Causes of quality deviations in design and construction. Journal of
Construction Engineering and Management, 118: 34-49.
Fazio, P., Moselhi, O., Théberge, P. and Revay, S. 1988. Design Impact of Construction
Fast-Track. Construction Management and Economics, 5: 195-208.
Ford, D. 1995. The Dynamics of Project Management — An Investigation of the Impacts of
Project Process and Coordination on Performance. Dissertation. Massachusettes Institute of
Technology, Cambridge, MA, USA.
Ford, D. N. and Sterman, J. D. 1998. Dynamic Modeling of Product Development Processes.
System Dynamics Review, 14: 31-68.
Ford, A and Flynn, H. 2005. Statistical Screening of System Dynamics Models. Systems
Dynamics Review, 21: 273-303.
Hwang, B., Thomas, S., Haas, C. and Caldas, C. 2009. Measuring the Impact of Rework on
Construction Cost Performance, Journal of Construction Engineering and Management, 135:
187-198.
Josephson, P. 1999. Causes and costs of defects in construction a study of seven building
projects. Automation in Construction, 8: 681-687.
Love, P.E.D., and Heng, L. 2000. Quantifying the causes and costs of rework in construction.
Construction Management and Economics, 18: 479-490.
Nelson. P. Repenning. 2001. Understand Fire Fighting in New Product Development, Journal of
Product Innovation Management, 18: 265-300.
Pefia-Mora, F. and Li, M. 2001. Dynamic Planning and Control Methodology for Design/Build
Fast-Track Construction Projects. Journal of Construction Engineering and Management, 127:
1-17.
Smith, R. P. and Eppinger, S. D. 1997. Identifying Controlling Features of Engineering Design
Iteration. Management Science, 43:276-293.
Sterman, J. D. 2000. Business Dynamics — Systems Thinking and Modeling for a Complex World ,
McGraw Hill, New York, NY, USA.
Taylor, T. and Ford, D. 2006. Tipping Point Failure and Robustness in Single Development
Projects. Systems Dynamics Review, 22: 51-71.
Taylor, T., Ford, D., and Ford, A. 2007. Model analysis using statistical screening: Extensions
and example applications. Proceedings of the 25" International Conference of the System
Dynamics Society. July 29-August 2. Boston, MA, USA.
Taylor, T. and Ford, D. 2008. Managing Tipping Point Dynamics in Complex Construction
Projects. Journal of Construction Engineering and Management, 134(6), 421-431.
Taylor, T. and Ford, D. 2010. Improving Model Understanding Using Statistical Screening.
System Dynamics Review, 26: 73-87.
15 of 15
